Research On Technology Of Decreasing False Alarms Of Mechatronics BIT At Feature-Level

Built-in test (BIT) technique is an important approach to improve testability and diagnostic capability of equipments and devices greatly. The domain of BIT application has extended from electronic equipments to mechantronic equipments. However, the high false alarm rate (FAR) is one of the important factors that prevent BIT from being more extensively applied. How to reduce the high FAR and retain high fault detection rate (FDR) is a key issue of BIT that should be well solved.Before most research of false alarm problem attached importance to enhance fault classifier diagnosis capability, and ignored the fault feature information quality of mechatronics system. Feature information quality of feature level would directly influence BIT diagnosis result. In this paper, theory for reasons of false alarm (FA) and technologies for reducing FAR at feature level of BIT systems are studied.The present research at feature level has two main issues. One is absent research of the mechanism of their influences of feature information on the BIT. The other is absent research of decreasing FA technologies especially fit to mechatronics BIT technologies. Supported by Research on Design Technology of Mechatronics BIT, and taking the high BIT FAR into account, this paper is aiming to solve the above two issues of the BIT at feature level. The content of the dissertation is as follows. 1. The reasons and mechanism of how the feature information quality influences the BIT performance are analyzed systematically.In order to analyze how the feature level influence the BIT performance, the factors that induce the feature information quality are analyzed first, and the mechanisms of feature level influencing FA are investigated. The quantitative analysis and simulation results show feature irrelevancy to fault condition, feature redundancy and feature unmatched to fault classifier is the cause of false alarm. Three formulas for feature level to ensure low FAR are attained.2. The technology of feature level to reduce BIT false alarms is studied deeply. (1) In order to solve FA brought by the mechatronics BIT nonlinear feature redundancy, the feature extraction based on the kernel principal component analysis (KPCA) methods are presented in view of merit of KPCA that has the ability to deal with nonlinear features. Because kernel function play important role in KPCA, a kernel function optimization method based on matrix similarity is proposed. The method first give a fine Gram matrix of kernel function out, then search an approximate kernel function in a finite range based on the matrix similarity. KPCA using optimization kernel...